Deep learning for automatic segmentation of thigh and leg muscles

Area of Science:

• Medical imaging analysis
• Deep learning in radiology
• Quantitative muscle imaging

Background:

• Accurate segmentation of muscle regions of interest (ROIs) is crucial for quantitative imaging in neuromuscular diseases.
• Manual ROI segmentation is time-consuming and lacks standardization.
• Existing automatic methods show promise but require further refinement.

Purpose of the Study:

• To develop a supervised deep learning approach for automatic segmentation of thigh and leg muscles.
• To create a unified framework for robust ROI segmentation.
• To improve the efficiency and standardization of muscle imaging analysis.

Main Methods:

• A supervised deep learning network was developed for muscle segmentation.
• The approach utilizes a unified framework for region of interest (ROI) delineation.
• The method was trained and validated on manually segmented images.

Main Results:

• The deep learning network achieved high segmentation accuracy, with Dice Scores between 0.89 and 0.95.
• The method demonstrated strong performance in segmenting muscles affected by varying degrees of fatty replacement.
• Accurate segmentation was achieved for both mild and severe disease manifestations.

Conclusions:

• The developed deep learning method offers a promising solution for automatic muscle segmentation.
• The approach shows potential for translation to other skeletal muscle groups.
• The method is adaptable to various MRI sequences with different resolutions and contrasts.